Convex Optimization and Applications in Classification Problems

Professor Laurent El Ghaoui

ABSTRACT

The success of Support Vector Machines is a witness to the strong interplay between convex optimization and classification problems.  In this talk, we explore several classification problems that can be formulated, and solved as, convex optimization problems.  Topics will include: SVMs and robustness with respect to input data; probabilistic versions of SVMs, such as minimax probability machines and entropy-constrained SVMs; kernel optimization using semidefinite programming.